1. Matt Chaput
  2. whoosh
  3. Issues
Issue #359 new

<type 'exceptions.AttributeError'>: 'DATETIME' object has no attribute 'numtype'

Carlos Sanchez
created an issue

in the NUMERIC class, the setstate method calls _min_max which refers to the field numtype which has not been loaded at the time. This does not allow loading of indices created with version 2.5.2 or earlier

I am attaching a index created w/ version 2.5.2 along w/ a definition of a custom field (VIDEO_TIMESTAMP) namespace (nti.contentsearch)

def setstate(self, d): self.dict.update(d) self._struct = struct.Struct(">" + self.sortable_typecode) if "min_value" not in d: d["min_value"], d["max_value"] = self._min_max()

Comments (7)

  1. Matt Chaput repo owner

    Unfortunately I can't load your index because of references to external modules. However, I can't figure out how it could be numtype, since numtype should always be in the pickled object's dictionary. Could you please paste in the error and traceback you're getting?

    As a workaround, since you have the schema available as a function, you can pass when you open the index:

    ix = index.open_dir("index", "vtrans_CLC3403_LawAndJustice", schema=schema)

    (Or in Storage.open_index() if that's how you're opening the index.) The code will skip the pickled schema and use the one you passed in.

  2. Carlos Sanchez reporter

    This is the exception I am getting

    File "/Users/csanchez/VirtualEnvs/nti.dataserver/lib/python2.7/site-packages/Whoosh-2.5.3-py2.7.egg/whoosh/index.py", line 136, in exists_in ix = open_dir(dirname, indexname=indexname) File "/Users/csanchez/VirtualEnvs/nti.dataserver/lib/python2.7/site-packages/Whoosh-2.5.3-py2.7.egg/whoosh/index.py", line 123, in open_dir return FileIndex(storage, schema=schema, indexname=indexname) File "/Users/csanchez/VirtualEnvs/nti.dataserver/lib/python2.7/site-packages/Whoosh-2.5.3-py2.7.egg/whoosh/index.py", line 421, in init TOC.read(self.storage, self.indexname, schema=self._schema) File "/Users/csanchez/VirtualEnvs/nti.dataserver/lib/python2.7/site-packages/Whoosh-2.5.3-py2.7.egg/whoosh/index.py", line 646, in read schema, segments = loader(stream, gen, schema, version) File "/Users/csanchez/VirtualEnvs/nti.dataserver/lib/python2.7/site-packages/Whoosh-2.5.3-py2.7.egg/whoosh/legacy.py", line 61, in load_110_toc schema = ru.load() File "/opt/local/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/pickle.py", line 858, in load dispatchkey File "/opt/local/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/pickle.py", line 1217, in load_build setstate(state) File "/Users/csanchez/Documents/workspace/NextThoughtPlatform/nti.dataserver/src/nti/contentsearch/_whoosh_schemas.py", line 149, in setstate super(VIDEO_TIMESTAMP, self).setstate(d) File "/Users/csanchez/VirtualEnvs/nti.dataserver/lib/python2.7/site-packages/Whoosh-2.5.3-py2.7.egg/whoosh/fields.py", line 514, in setstate d["min_value"], d["max_value"] = self._min_max() File "/Users/csanchez/VirtualEnvs/nti.dataserver/lib/python2.7/site-packages/Whoosh-2.5.3-py2.7.egg/whoosh/fields.py", line 517, in _min_max numtype = self.numtype ConfigurationExecutionError: <type 'exceptions.AttributeError'>: 'VIDEO_TIMESTAMP' object has no attribute 'numtype'

  3. Carlos Sanchez reporter

    The class in question is defined as


    def _parse_datestring(self, qstring):
        result = videotimestamp_to_datetime(qstring)
        return result
    def __setstate__(self, d):
        # d['bits'] = 64
        # d['numtype'] = int
        super(VIDEO_TIMESTAMP, self).__setstate__(d)

    and the schema is

    fields.Schema(containerId=fields.ID(stored=True, unique=False), videoId=fields.ID(stored=True, unique=False), language=fields.ID(stored=True, unique=False), keywords=fields.KEYWORD(stored=True), start_timestamp=VIDEO_TIMESTAMP(stored=True), end_timestamp=VIDEO_TIMESTAMP(stored=True), last_modified=fields.DATETIME(stored=True))

    As you can see I worked around the issue by simply setting the bit and numtype fields. The index was created using 2.5.2 and read in by 2.5.3

  4. Matt Chaput repo owner

    The weird thing is I can see the numtype in the pickle string. I'm not sure what's going on here. I'll work on faking on your environment to try to make the pickle loadable so I can reproduce the error.

  5. Matt Chaput repo owner

    I got the index to load (using whoosh.util.loading.RenamingUnpickler to redirect nti.contentsearch._whoosh_schemas.VIDEO_TIMESTAMP to a copy of the code you gave above) and it worked for me :(

  6. Carlos Sanchez reporter

    I am surprised to that those two fields are not in pickled in the schema. Well I have a workaround Thanks for all the effort. I am going to recreate my indices with 2.5.3

  7. Log in to comment